Witsenburg et al. Journal 2012, 11:53 http://www.malariajournal.com/content/11/1/53

RESEARCH Open Access The evolutionary host switches of Polychromophilus: a multi-gene phylogeny of the malaria genus suggests a second invasion of by a haemosporidian parasite Fardo Witsenburg1*, Nicolas Salamin1,2 and Philippe Christe1

Abstract Background: The majority of infect birds or reptiles, but many important genera, including , infect mammals. Dipteran vectors shared by avian, reptilian and mammalian Haemosporida, suggest multiple invasions of Mammalia during haemosporidian evolution; yet, phylogenetic analyses have detected only a single invasion event. Until now, several important -infecting genera have been absent in these analyses. This study focuses on the evolutionary origin of Polychromophilus, a unique malaria genus that only infects (Microchiroptera) and is transmitted by bat flies (Nycteribiidae). Methods: Two species of Polychromophilus were obtained from wild bats caught in Switzerland. These were molecularly characterized using four genes (asl, clpc, coI, cytb) from the three different genomes (nucleus, apicoplast, mitochondrion). These data were then combined with data of 60 taxa of Haemosporida available in GenBank. Bayesian inference, maximum likelihood and a range of rooting methods were used to test specific hypotheses concerning the phylogenetic relationships between Polychromophilus and the other haemosporidian genera. Results: The Polychromophilus melanipherus and Polychromophilus murinus samples show genetically distinct patterns and group according to species. The Bayesian tree topology suggests that the monophyletic clade of Polychromophilus falls within the avian/saurian clade of Plasmodium and directed hypothesis testing confirms the Plasmodium origin. Conclusion: Polychromophilus’ ancestor was most likely a bird- or reptile-infecting Plasmodium before it switched to bats. The invasion of mammals as hosts has, therefore, not been a unique event in the evolutionary history of Haemosporida, despite the suspected costs of adapting to a new host. This was, moreover, accompanied by a switch in dipteran host. Keywords: Polychromophilus, Malaria, Haemosporida, Chiroptera, Plasmodium, Host switch, Phylogenetic analysis, Outgroup selection

Background Ceratopogonidae and Nycteribiidae respectively, while the Five genera belonging to the order of Haemosporida (Api- vectors of Nycteria and Rayella are unknown [1,2]. Culici- complexa) are known to infect mammals: Plasmodium, dae and Ceratopogonidae also act as vectors of the avian , Polychromophilus, Nycteria and Rayella and saurian Haemosporida [1,3]. These shared vectors [1,2]. The dipteran vectors of the first three haemospori- suggest that haemosporidian parasites might have invaded dian genera are represented by Culicidae (Anopheles spp.), mammals multiple times during their evolution. On the other hand, the switch to mammals is thought to have * Correspondence: [email protected] been an evolutionary demanding process for the parasite 1Département d’Ecologie et Evolution, Université de Lausanne, Biophore, [4] and therefore a rare event [5]. UNIL-Sorge, 1015 Lausanne, Switzerland Full list of author information is available at the end of the article

© 2012 Witsenburg et al; BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Witsenburg et al. Malaria Journal 2012, 11:53 Page 2 of 10 http://www.malariajournal.com/content/11/1/53

Molecular phylogenetic studies to date have been able hypotheses on its phylogenetic relationships: 1) it is to detect only a single host switching event to mammals: monophyletic with the mammalian Plasmodium/Hepato- mammalian Plasmodium and Hepatocystis,themain cystis clade with which it shares the vertebrate host type, mammal-infecting genera, had a common origin and or 2) it shares its most recent common ancestor with formed a monophyletic sister clade to sauropsid Plasmo- the subgenus (Haemoproteus), which has dium [6,7]. However, these phylogenetic studies suffer a similar vector. The genus Haemoproteus contains two from incomplete taxon sampling with most investigations avian subgenera which have different vectors. H. (Para- including, besides the genera Plasmodium and Hepato- haemoproteus) spp. use biting midges as vectors, and H. cystis, only the avian Haemoproteus and . (Haemoproteus) spp. are transmitted by Hippoboscidae, Consequently, with no knowledge of the evolutionary ori- whose closest relatives are the bat flies [17]. A phylo- gin of the other mammalian haemosporidian groups (i.e. geny based on ultrastructure and life-history traits Rayella, Nycteria, Polychromophilus), a second move into grouped Polychromophilus together with both subgenera mammals cannot be excluded. of Haemoproteus [8]. However, two recent molecular One possible approach for resolving this standing ques- phylogenetic studies based on part of the cytochrome b tion is to select a haemosporidian genus that could poten- sequence both suggest, despite their different topologies, tially have switched to mammal hosts independently of a close relationship between Polychromophilus and saur- mammalian Plasmodium/Hepatocystis. A good candidate opsid Plasmodium [18,19]. This fact provides a third genus for this is Polychromophilus as it is well described, hypothesis: 3) Polychromophilus is monophyletic with with the majority of its life cycle well documented, includ- sauropsid Plasmodium (see Figure 2). ing its vector stage. Moreover, it infects mammals but is The aim of this study was to test these three hypoth- not transmitted by Culicidae like Plasmodium, nor Cerato- eses against each other. Though previous studies on the pogonidae like Hepatocystis, but by Nycteribiidae (Diptera: phylogenetic relationships of Polychromophilus have Hippoboscoidea). Furthermore, Polychromophilus’ verte- been done, all used only a single gene. Different genes in brate host species range is restricted to the insectivorous a single organism can show different evolutionary pat- bats (Microchiroptera). A phylogenetic analysis of Poly- terns and it is, therefore, recommended to use multiple chromophilus can, therefore, elucidate whether it arose genes for accurate relationship estimation [20]. The four through an independent switch to mammal hosts [8]. genes from three different genomes sequenced for this Only five species of Polychromophilus are known to study represent two species of Polychromophilus (being exist. While they can be distinguished by their slight differ- the two type species of the two proposed subgenera). ences in ultrastructure, they are mainly classified based on These sequences were subsequently combined with an host-type [9,10]. Landau et al. [10] proposed dividing the existing dataset of 60 species of Haemosporida to clarify genus into two subgenera based on their gametocyte mor- the phylogenetic relationships and gain insight into the phology: 1) the subgenus Polychromophilus,withP. (P.) evolutionary host switches of Polychromophilus. melanipherus as the type species, which has gametocytes similar to the type ‘malariae’; 2) the subgenus Bioccala Methods with type species Polychromophilus (B.) murinus whose Sample collection and preparation gametocytes resemble the benign tertian parasites of birds Four Miniopterus schreibersii (Schreibers’ bent-winged and reptiles (see Figure 1) [10]. Later, it was even sug- bat) were caught using mist nets in the entrance of an gested that the subgenus Bioccala be raised to genus level abandoned mine in western Switzerland under authoriza- [11]; however, this was not reflected in the literature [12]. tion 2203 issued by the Veterinarian Service of Canton Moreover, the morphological distinctions between the Vaud, Switzerland. Blood was obtained by puncturing the species have been described as ‘slight’ [9] and how well uropatagial vein with a 0.5 mm gauge needle (Neolus). they reflect the genetics of the genus has not been studied. The blood beads that consequently formed on the uropa- The Nycteribiidae vectors are also known as nycteribids tagium (between 10 and 30 μl total) were taken up in a or bat flies. These haematophagous flies are completely microvette with EDTA (Sarstedt) and stored at 4°C until adapted to a parasitic lifestyle in the fur of bats in that further analysis. Haemostatic cotton was applied on the they have lost their wings, have no or reduced eyes and punctured vein until the bleeding had stopped before possess hooking claws which allow them swift movements releasing any bats. through the fur [13,14]. Coradetti [15] was the first to One drop of blood was applied to a glass microscope detect sporozoites in their salivary glands and later studies slide for later visual identification of the parasite species. confirmed his finding [12,16]. After smearing the blood, the slide was dried and imme- When an evolutionary conservation of hosts is diately submerged in 100% methanol for fixation. Finally assumed, Polychromophilus’ unique host-vector combi- 5% Giemsa-staining was applied for one hour to stain the nation of Mammalia and Nycteribiidae gives rise to two cells. DNA was extracted from whole blood using the Witsenburg et al. Malaria Journal 2012, 11:53 Page 3 of 10 http://www.malariajournal.com/content/11/1/53

Figure 1 Polychromophilus murinus and P. melanipherus gametocytes.a.Polychromophilus murinus infection isolated from blood of a Daubenton’s bat; b. P. melanipherus infection isolated from blood of a Schreibers’ bent-winged bat. (Thin blood film, Giemsa-staining, phase contrast image).

DNeasy Blood and Tissue kit (Qiagen). Megali et al. [18] sequence from the apicoplast, caseinolytic protease C provided extracted DNA samples from blood of Myotis (clpc, 502 bp); and one nuclear DNA sequence, adenylo- daubentoni (Daubenton’s bat) which contained P. muri- succinate lyase (asl, 186 bp). nus infections. These infections were previously shown to All primer pairs used for the polymerase chain reactions be characterized by different cytochrome b haplotypes were taken from Martinsen et al. [6] with the exception of [18]. coI nested Po, which was designed during this study (see Table 1 for primer sequences). All reactions started with Molecular analysis an initial denaturation phase at 94°C for four minutes and For the phylogenetic reconstruction, four genes were one minute for the first and nested PCRs, respectively. selected from the three cellular genomes: two mitochon- The reactions ended with an annealing phase at 72°C for drial DNA sequences, cytochrome b (cytb, 607 bp) and 7 min. All cycles started for 30 s at 94°C, but the other cytochrome oxidase subunit I (coI, 768 bp); one DNA cycle conditions and the number of cycles differed Witsenburg et al. Malaria Journal 2012, 11:53 Page 4 of 10 http://www.malariajournal.com/content/11/1/53

Figure 2 The hypothetical phylogeny of the genus Polychromophilus and the other Haemosporida. The hypothetical branches are marked in orange and based either on the conservation of the vertebrate host (hypothesis 1), the conservation of the dipteran vector (hypothesis 2), or based on previous molecular studies of the cytb gene (hypothesis 3). depending on the primer pair used (see Table 1). The 25 MgCl (except for the reactions with the coI primers, which μl reaction volume contained 3 μl of extraction product, had a total of 3 mM MgCl). The nested PCR reaction 0.25 U Taq polymerase, 0.3 mM of both primers, 0.25 volume was similar except for the extraction product, mM dNTP’s, 1× Qiagen PCR buffer and a total of 2 mM which was replaced with 1 μl of product of the first PCR.

Table 1 Name, sequence and PCR conditions of the primers used Name Primer sequence Annealing Extension Cycles asl/outer fw GSKAARTTTAATGGKGCTGTWGG 47°C, 30 s 72°C, 50 s 35 rv GGATTAAYTTTATGAGGCATTG asl/nested fw GCTGATMAAAATRTTGATTGG 50°C, 30 s 72°C, 30 s 38 rv GAGGCATTGTACTACTWCC clpc/outer fw AAACTGAATTAGCAAAAATATTA 50°C, 30 s 72°C, 50 s 38 rv CGWGCWCCATATAAAGGAT clpc/nested fw GATTTGATATGAGTGAATATATGG 48°C, 30 s 72°C, 30 s 40 rv CCATATAAAGGATTATAWG coI/outer fw CTATTTATGGTTTTCATTTTTATTTGGTA 57°C, 30 s 72°C, 50 s 35 rv AGGAATACGTCTAGGCATTACATTAAATCC coI/nested Po fw AGCAATATCAATAGCTGCATTACCT 62°C, 30 s 72°C, 50 s 38 rv GATTTTCTTCAATATAATGCCTGGA cytb/outer fw TAATGCCTAGACGTATTCCTGATTATCCAG 55°C, 30 s 72°C, 50 s 35 rv TGTTTGCTTGGGAGCTGTAATCATAATGTG cytb/nested fw TCAACAATGACTTTATTTGG 55°C, 30 s 72°C, 50 s 40 rv TGCTGTATCATACCCTAAAG All denaturation and final extension periods are the same for all primer-pairs. Witsenburg et al. Malaria Journal 2012, 11:53 Page 5 of 10 http://www.malariajournal.com/content/11/1/53

For the asl amplification the first PCR product was puri- Rooting the tree fied, which resulted in a better performance of the nested Which outgroup to use has been a matter of debate lately. reaction. Perkins and Schall [7] identified Leucocytozoon as the All successfully amplified samples were purified accord- most primitive clade of the order, using as an ing to the manufacturer’s protocol using the Wizard PCR outgroup in their analysis of cytb sequences. But a recent Clean-Up system (Promega) or the Minelute PCR Purifica- study by Outlaw and Ricklefs [19] demonstrated that, tion kit (Qiagen) in the case of asl. DNA concentrations when using a relaxed molecular clock, Leucocytozoon were estimated by visualization on a 1.5% agarose gel with becomes the most derived group, effectively turning the a 100 bp reference ladder (Roche). For the sequencing tree inside-out. The authors argue that most ancient diver- reactions ~20 ng of purified PCR product, 2 μlBigDye gence should be between the mammal-infecting Plasmo- Terminator v3.1 and 1 μlof10mMprimerweremixedto dium and Hepatocystis on the one side, and avian/saurian a10μl volume. Sequence analysis was performed on an Plasmodium, both subgenera of Haemoproteus and Leuco- ABI Prism 3,100 genetic analyser (Applied Biosystems). cytozoon on the other. Sequence chromatographs were checked for ambiguities For the phylogenetic tree reconstructions, the Leucocy- with Chromas Lite v2.01 (Technelysium). tozoon spp. were initially selected as the outgroup, but these results were tested for their robustness by redoing Phylogenetic reconstruction the analyses using different rooting methods: 1) forcing The obtained sequence data were combined with the same the mammalian Plasmodium/Hepatocystis clade as out- gene sequences of 60 other haemosporidian species group instead of Leucocytozoon; 2) adding amino acid obtained from GenBank (see Additional file 1). These 60 sequences of the more distantly related spp. as species represent the major clades of the Haemosporida, the outgroup (see Additional file 1) and repeating the ML i.e. Leucocytozoon, Haemoproteus (Haemoproteus), analyses; 3) using the molecular clock methods similar to Haemoproteus (Parahaemoproteus), Hepatocystis and Plas- Outlaw and Ricklefs [19] but with varying priors: a Yule modium (including mammalian, avian and saurian). or birth-death tree prior, a strict, a log-normal relaxed or Sequences were aligned with ClustalW as implemented in an exponential relaxed clock with a GTR + Γ + I substi- MEGA version 5 [21]. The single-gene alignments were tution model, 20 million generations sampling every concatenated using FASconCAT [22]. 2,000 generations and two independent MCMC runs All phylogenetic reconstructions were done using both using BEAST [27-29]. Maximum Likelihood (ML) analysis and Bayesian infer- ence (BI). For ML analysis, the PhyML software [23] was Topological tests used for the single-gene alignments. Since PhyML does The obtained Bayesian majority rule consensus tree was not allow for partitioning of the data RAxML [24] was compared with each of the four Bayesian single-gene used for the concatenated alignment. Models of nucleotide majority rule consensus trees to rule out any conflict in substitution were GTR + Γ + I for cytb, co1 and clpc and topology. The Kishino-Hasegawa tests [30] were performed GTR + Γ for asl, as determined by MrAIC [25]. For each in Treefinder [31]. The tests proved non-significant for all analysis, the transition rates of the GTR model, the shape genes but asl (see Table 2). This gene was, therefore, of the Γ-distribution and the proportion of invariable sites removed from the concatenated alignment and a new phy- were estimated by the program. Both the RAxML and logenetic reconstruction was performed on the remaining PhyML analyses were assessed by performing 1,000 boot- genes only. strap replicates. For the Bayesian analysis, the same models of character Results evolution as described for the ML analyses were imple- The stained slides showed erythrocytes infected with mented with MrBayes 3.1.2 [26]. In the concatenated ana- slightly oval-shaped gametocytes (see Figure 1b). The lysis the data were again partitioned by gene, where each granular appearance and pinkish staining at the nucleus fit partition had its corresponding model and independent the description of Polychromophilus melanipherus as given parameter estimations. The MCMC algorithm was done by Garnham [1]. The morphology of the observed gameto- with four chains and was run for 20,000,000 generations, cytes could therefore be linked to the molecular sequences sampling every 1,000 generations. Two independent runs obtained from the infections (Table 3). were performed to assess convergence to the correct pos- Although the reasons for this DNA region to be rejected terior distribution. All parameters were checked for con- by the topology tests are unclear, the length of the asl vergence using Tracer v1.5 and the first 10% of samples of gene fragment sequenced in this study is very small (186 each run was discarded as burn-in. All computations were bp). This could suggest that random errors are responsible performed on the Vital-IT cluster of the Swiss Institute of for creating the incongruences observed with this gene. Bioinformatics. Adding other, and especially longer, nuclear genes would Witsenburg et al. Malaria Journal 2012, 11:53 Page 6 of 10 http://www.malariajournal.com/content/11/1/53

Table 2 The Kishino-Hasegawa topological test results the two species of Polychromophilus form well sup- Single gene tree Concatenated 4 genes tree ported separate sister clades (see Figure 3). The distinc- lnL lnL pKH tion between P. melanipherus and P. murinus has often asl -3561.406 -3642.361 < 0.001* been made based on host species, since P. melanipherus clpc -7718.852 -7728.157 0.2696 was by definition confined to Miniopterus schreibersii as coI -10062.16 -10074.33 0.2336 hosts. This distinction, however, has been qualified as ‘ ’ ‘ ’ cytb -6856.617 -6859.401 0.4323 arbitrary and unsatisfactory [1]. This study demon- For each gene the likelihood of the phylogeny of that gene was compared to strates for the first time that there is a clear genetic dis- the phylogenetic reconstruction based on all four genes. The log-likelihood tinction between the two Polychromophilus species, values and p-values are shown per gene alignment. Only the asl alignment confirming their taxonomic status of different species gives a significantly worse likelihood value for the tree based on the combined data, which indicates conflicting topologies. from a molecular point of view. For each of the three hypotheses on the Polychromophilus origin a However, to determine whether this level of genetic corresponding topology was constructed. This was done by restricting the divergence between P. murinus and P. melanipherus mer- placing of Polychromophilus during tree reconstruction in RAxML, forcing it either with the mammal-infecting Plasmodium/Hepatocystis clade (hypothesis its their placement in different subgenera [10] or even dif- 1), the Haemoproteus (Haemoproteus) clade (hypothesis 2) or with the sauropsid ferent genera [11], other species of the genus should be Plasmodium clade (hypothesis 3). These restricted topologies were then tested together with the topology produced by the maximum likelihood analysis added (e.g. P. deanei [35] and P. adami [10]). Without using a Shimodaira-Hasegawa test [32] as implemented in PAML 4 [33]. these supplementary species, the overall observed genetic diversity within the genus Polychromophilus is low; it is certainly bring more information to test if the evolutionary clearly less than that of the genera Plasmodium and Hae- relationships estimated from the different genomes are moproteus or even less than the diversity found in subge- congruent or if specific gene trees best represent the evo- nera like P. (Vinckeia) and H. (Parahaemoproteus).No lution of each DNA regions. Different cellular genomes critical level of genetic diversity exists as a precondition often have different evolutionary histories; even within a for the elevation of a subgenus, but the low diversity single genome not all genes show the same phylogenetic found within Polychromophilus does suggest that confirm- relationships [20]. ing P. (Bioccala) as a separate genus would cause a taxo- Figure 3 presents the reconstructed phylogenetic trees nomic asymmetry within the Haemosporida. using the combined data of cytb, coI and clpc by ML Two more haemosporidian genera infecting bats have and BI. The analyses produce no conflict on any of the been described: Dionisia [36] and Biguetiella [11]. Both major nodes. All major genera and subgenera are recov- contain only a single species and are described as ‘little dif- ered and represented in the phylogenetic tree by sepa- ferent’ from Polychromophilus (Polychromophilus) [36] and rate monophyletic clades, with the exception of the as ‘a vicariant form of’ Polychromophilus (Bioccala) [11], sauropsid Plasmodium clade, which contains Polychro- respectively. Whether their similarities to Polychromophi- mophilus within it. lus spp. are because of convergence or shared ancestry can only be tested by combining the morphological data with Diversity of Polychromophilus species molecular methods [37]. A big obstacle in studying these Polychromophilus forms its own clearly defined clade in unfamiliar species however is the lack of observations. No both the BI and ML reconstructions. Within this clade, other records of Biguetiella or Dionisia exist. Single

Table 3 The haplotypes and corresponding accession numbers for GenBank per sequenced sample per gene asl clpc coI cytb ind. ht. acc. nb. ht. acc. nb. ht. acc. nb. ht. acc. nb. 104 Pmu1 JN990725 Pmu1 JN990723 Pmu1 JN990718 Pmu1 JN990712 114 Pmu1 .. Pmu2 JN990724 - - Pmu1 .. 156 Pmu1 .. Pmu1 .. Pmu2 JN990719 Pmu2 JN990713 A2111 - Pme3 JN990720 Pme3 JN990714 Pme3 JN990708 A2112 Pme2 JN990726 Pme4 JN990721 Pme4 JN990715 Pme4 JN990709 A2113 - - Pme5 JN990722 Pme5 JN990716 Pme5 JN990710 A2114 - - - - Pme6 JN990717 Pme6 JN990711 Samples 104, 114 and 156 are Polychromophilus murinus, sampled from Myotis daubentoni and shared some haplotypes. The samples A2111-A2114 are Polychromophilus melanipherus from Miniopterus schreibersii and never shared haplotypes. For each unique haplotype, the GenBank accession number is mentioned only once in the table. ‘..’: accession number already mentioned. ‘-’: sequencing was unsuccessful. None of the topologies obtained by independent analyses of the separate genes conflicted with the topology resulting from the concatenated alignment (Kishino-Hasegawa tests: cytb: Δlnl = 2.8, pKH = 0.432, coI: Δlnl = 12.1, pKH = 0.234, clpc: Δlnl = 9.3, pKH = 0.270), except for asl (Δlnl = 81.0, pKH < 0.001). Despite this strong rejection, both the ML and BI trees of asl had only few supported nodes and only closely related pairs were recovered (data not shown). A possible cause of the incongruence detected could be positive selection events in the evolution of the asl nuclear sequence [34]. However, analyses performed with Codeml [33] did not show signs of positive selection on the nuclear gene. Witsenburg et al. Malaria Journal 2012, 11:53 Page 7 of 10 http://www.malariajournal.com/content/11/1/53

Figure 3 Polychromophilus shares its most recent common ancestor with avian and reptilian Plasmodium. Shown is the 50% majority- rule consensus tree from the Bayesian inference analysis. The phylogenetic reconstruction using maximum likelihood produced a similar tree. For clarity all clades except the Polychromophilus are collapsed and replaced by coloured triangles. Each colour represents a different haemosporidian group. The dots indicate Bayesian node support. Closed dots indicate a posterior probability ≥ 0.95, open dots a posterior probability ≥ 0.90. Node values indicate bootstrap values. Branch lengths represent the number of substitutions. The single blue branch belongs to a Plasmodium sp. infecting the skink Egernia stokesii. descriptions of new parasite species found in a limited 0.87 and 1. However the root itself does change depending number of hosts are a problem encountered more often on the prior set. The Yule and log-normal prior lead to by parasitologists and can severely hamper classification the placement of the Leucocytozoon as the outgroup, [37]. whereas the mammalian Plasmodium/Hepatocystis clade is placed as the outgroup with the birth-death tree and Polychromophilus’ placement in the phylogeny of exponential relaxed clock (see Additional file 2). Haemosporida The ML analysis using Babesia as an outgroup pro- The bootstrap value (69/100) suggests that the Polychro- duces a topology with very little support. All major nodes mophilus clade is restricted to the Plasmodium branch of have bootstrap values of < 50/100, so no outgroup can be the haemosporidian tree. Even though this node also appears in the Bayesian majority rule consensus tree, the support for it is actually very weak (posterior probability = Table 4 The Shimodaira-Hasegawa topological test 0.73). However, the alternative hypothesis 2, that Polychro- results comparing the three hypothetical topologies mophilus shares its most recent common ancestor with Tree lnL pSH the subgenus Haemoproteus (Haemoproteus), is clearly Best tree (hypothesis 3) -25126.753 - rejected (Shimodaira-Hasegawa test; see Table 4). Hypothesis 1 -25130.147 0.578 Most of the alternative rooting methods favour hypoth- Hypothesis 2 -25154.740 0.023* esis 3. Indeed, rooting the tree with the mammalian Plas- The best tree was the tree provided by the maximum likelihood analysis (see modium clade instead of Leucocytozoon,assuggestedby Figure 3) and concurred with hypothesis 3. The log-likelihoods of the other two trees, based on hypothesis 1 and 2 (see Figure 2), are compared with the best Outlaw and Ricklefs [19], validates the conclusion of a tree. The hypothesis 2 tree, which has Polychromophilus grouped with sauropsid origin of Polychromophilus in both BI and ML Haemoproteus, has a significantly worse fit and can be rejected. It is less clear where within the Plasmodium clade Polychromophilus belongs. (see Additional file 2). Neither phylogenetic method indicates that Polychromophilus originated from The choice of the prior distributions guiding either the mammalian Plasmodium/Hepatocystis and both instead produced topologies distribution of mutation rates across the tree (log-normal suggesting a sauropsid origin (see Figure 3). However, the actual support for the node separating the mammalian clade from sauropsid Plasmodium/ vs exponential) or the divergence times (Yule vs birth- Polychromophilus clade is low. The BI supports the monophyly of sauropsid death) does not change the conclusion. All the molecular Plasmodium and Polychromophilus (hypothesis 3) with a posterior probability of 0.92, but the ML support of that same critical node is absent (bootstrap value clock analyses place Polychromophilus within the saurop- of 40/100). The topological test comparing the different phylogenetic scenarios sid Plasmodium clade, with clade credibilities between did not provide more support for either hypothesis 1 or 3 (see Table 4). Witsenburg et al. Malaria Journal 2012, 11:53 Page 8 of 10 http://www.malariajournal.com/content/11/1/53

identified, nor can Polychromophilus be placed within the big obstacle in apicomplexan research as a whole [42] and tree with any confidence (see Additional file 3). The haemosporidian research in particular [43]. Therefore, genetic divergence between Babesia () and caution is required when naming species for GenBank. the Haemosporida is very high, which results in a very long branch leading to the Babesia lineages. This changes Switch of host, switch of vector the rooting procedure to a problem of ‘long-branch Parasitizing a new, mammalian host likely necessitated attraction’ with all corresponding biases [38], and these many adaptive changes, given their characteristic, non- analyses should therefore be approached with caution nucleated red blood cells. The cytb DNA region sequenced [19]. here showed long branches of non-synonymous substitu- None of the used phylogenetic methods reject our tions separating the avian from the mammal clade [4]. third hypothesis, stating that Polychromophilus is mono- Many lineages have become extinct over time during the phyletic with the sauropsid Plasmodium clade. ML and evolution towards the mammalian and avian Plasmodium the topological test could not discriminate between lineages [44]. Nevertheless Polychromophilus’ origin sug- hypothesis 2 and 3, but BI and molecular clock rooting gests that the switch to mammalian hosts happened at methods gave more support for the latter hypothesis. least twice during Haemosporida evolution. Rayella is These analyses are far from conclusive, but do suggest thought to have originated from Hepatocystis [45] and has that Polychromophilus did not evolve from a mammal- been classified as such [1], but Nycteria’s origins are more infecting ancestor, but has instead invaded the mamma- elusive; whether it is a case of yet another independent lian class of hosts independently. host switch, or an ancient mammalian Plasmodium line- Our results show that the three DNA regions used in age that has survived the pruning on that branch, remains the combined matrix do not provide sufficient phyloge- to be investigated. netic information to unambiguously place the Polychromo- Haematophagy has appeared multiple times in the evo- philus lineage. When combining regions from different lution of the Diptera. It evolved once at the origin of the genomes, this could introduce sufficient conflict to reduce superfamily Hippoboscoidea and is shared by all its mem- the confidence in the reconstructed trees, even if the bers [17]. Consequently, many Hippoboscoidea spp. are topology tests did not identified major incongruence. The implicated in the transmission of diseases, most notably way forward to clearly place the Polychromophilus lineage sleeping sickness (Glossinidae) and malaria (Hippobosci- within the large Plasmodium clade is to sequence longer dae and Nycteribiidae). The relatively high relatedness of stretches of DNA regions, in particular from the nuclear the latter two families [17] is not reflected by their hae- genome, and to use gene tree approaches to identify the mosporidian parasites. This study convincingly rejected best evolutionary relationships at the species level [39,40]. the hypothesis that hippoboscid-transmitted H. (Haemo- proteus) shares its most recent common ancestor with the Previous findings nycteribid-transmitted Polychromophilus. A cospeciation The close relation between Polychromophilus and avian event of these Haemosporida with their dipteran hosts Haemosporida has been suggested before. Carreno et al. can, therefore, clearly be excluded. [8] produced a phylogeny based on life-history and ultra- Instead, Polychromophilus’ ancestor must have been vec- structure characters and concluded that Polychromophilus tored by of a member of the Culicidae, as are all modern is most closely related to Haemoproteus, a hypothesis Plasmodium species. Culicidae are one of the oldest mem- rejected by the current study. Megali et al. [18] used a bers of the Diptera, an order with a higher radiation of 705 bp cytb fragment and concluded that Polychromophi- species than all terrestrial vertebrates put together [46]. lus shared its closest common ancestry with avian Plasmo- The phylogenetic distance between Culicidae and Nycteri- dium. However, the base of their tree was not well biidae is one of the largest within the order [46], yet the resolved. The authors themselves recommended the use of adaptations required for this new vector were seemingly multiple genes. acquired in parallel to those required for the new mamma- Duval et al. [41] discussed bat Haemosporida, but never lian host. identified the species. However, their molecular analyses, Because the Nycteribiidae are completely specialised to again using only cytb, grouped their samples clearly with bats, the first appearance of Polychromophilus in bats sauropsid Plasmodium, leading to a similar conclusion as must have been mediated by either mosquitoes or via the the current study. In their paper, they cautiously did not hippoboscid flies. Many Culicidae spp. feed on both mam- name their collected species. However, the corresponding mals and birds readily, and within the Hippoboscidae, the sequences that are available in GenBank have been identi- host switch from mammals to birds has happened several fied as ‘Hepatocystis sp.’. Based on the work of Megali times [17]. Therefore, both could have been responsible et al. [18] it is very likely that part of those sequences are for the first transmission. However, once Polychromophi- actually Polychromophilus species. Misidentification is a lus’ ancestor was introduced in bats, adapting to the Witsenburg et al. Malaria Journal 2012, 11:53 Page 9 of 10 http://www.malariajournal.com/content/11/1/53

nycteribid vectors likely had large fitness advantages. Spe- like to give our gratitude to Emily Clark and two anonymous referees for helpful comments improving this manuscript. We thank the Swiss Institute cifically, the haematophagous lifestyle of both males and of Bioinformatics for their support and the availability of the Vital-IT facilities. females combined with their high prevalence on bats [13], This research was supported by grant 31003A_120479 from the Swiss and ease of moving between bat-hosts (unpublished obser- National Science Foundation. vations), make the Nycteribiidae an ideal vector for the Author details protozoan parasite. However, this same switch to Nycteri- 1Département d’Ecologie et Evolution, Université de Lausanne, Biophore, biidae also limited the potential range of Polychromophilus UNIL-Sorge, 1015 Lausanne, Switzerland. 2Swiss Institute of Bioinformatics, vertebrate hosts to the Chiroptera. Quartier Sorge, 1015 Lausanne, Switzerland. Authors’ contributions Conclusions FW collected the samples, carried out the molecular work, performed the phylogenetic analyses and drafted the manuscript. NS participated in the The phylogenetic reconstruction of three genes of Polychro- phylogenetic analyses. PC conceived the study, collected the samples and mophilus spp. demonstrates that the P. melanipherus and interpreted the results. All authors read and approved the final manuscript. P. murinus are clearly two genetically distinct species. Only Competing interests the addition of the other Polychromophilus spp. can validate The authors declare that they have no competing interests. the current division of Polychromophilus in separate subge- nera. Polychromophilus is clearly not related to Haemopro- Received: 21 November 2011 Accepted: 22 February 2012 teus (Haemoproteus). Instead Bayesian inference and Published: 22 February 2012 molecular clock outgroup free phylogenetic reconstructions References suggest that the Polychromophilus most likely had a bird- 1. Garnham PCC: Malaria Parasites and Other Haemosporidia. 1 edition. Oxford: or reptile-infecting Plasmodium ancestor. The switch to Blackwell Scientific Publications; 1966. 2. Dasgupta B: A new malarial parasite of the . Parasitology mammalian hosts would, therefore, not have occurred 1967, 57:467-474. once, but at least twice in the haemosporidian evolutionary 3. Valkiũnas G: Avian Malaria Parasites and Other Haemosporidia. Englishth past. This event was accompanied by the adaptation to a edition. Boca Raton: CRC Press; 2005. 4. Outlaw DC, Ricklefs RE: Comparative gene evolution in haemosporidian new, phylogenetically distant dipteran vector. () parasites of birds and mammals. Mol Biol Evol 2010, 27:537-542. Additional material 5. Yotoko KSC, Elisei C: Malaria parasites (Apicomplexa, Haematozoea) and their relationships with their hosts: is there an evolutionary cost for the specialization? J Zool Syst Evol Res 2006, 44:265-273. Additional file 1: Species name, host and accession numbers of 6. Martinsen ES, Perkins SL, Schall JJ: A three-genome phylogeny of malaria sequences retrieved from GenBank for the phylogenetic parasites (Plasmodium and closely related genera): evolution of life- reconstructions. This table contains additional host information and the history traits and host switches. Mol Phylogenet Evol 2008, 47:261-273. GenBank accession numbers of all genes used for the phylogenetic 7. Perkins SL, Schall JJ: A molecular phylogeny of malarial parasites analyses. Not all gene sequences are available for all species, missing recovered from cytochrome b gene sequences. J Parasitol 2002, sequences are denoted by ‘-’. 88:972-978. Additional file 2: Changing topologies for alternative rooting 8. Carreno RA, Kissinger JC, McCutchan TF, Barta JR: Phylogenetic analysis of methods. Changing topologies acquired by different methods of haemosporinid parasites (Apicomplexa: Haemosporina) and their phylogenetic reconstruction. Irrespective of the root, Polychromophilus coevolution with vectors and intermediate hosts. Arch Protistenkd 1997, remains nested within the sauropsid Plasmodium clade. a. The original 148:245-252. Polychromophilus best tree from maximum likelihood reconstruction, but now rooted with 9. Garnham PCC: The zoogeography of . Ann Parasitol Hum the mammalian Plasmodium/Hepatocystis, as suggested by Outlaw and Comp 1973, 48:231-242. Ricklefs [16]. Topologies b. and c. are acquired using a relaxed molecular 10. Landau I, Rosin G, Miltgen F, Hugot JP, Leger N, Beveridge I, Baccam D: Sur Polychromophilus clock with no predefined root, GTR + Γ + I substitution model, 20 le genre (, parasite de million generations sampling every 2,000 generations and two microchiroptères). Ann Parasitol Hum Comp 1980, 55:13-32. independent MCMC runs using BEAST. All nodes have clade credibilities 11. Landau I, Baccam D, Ratanaworabhan N, Yenbutra S, Boulard Y, > 0.5 b. Topology acquired with the Yule tree prior and an exponential Chabaud AG: Nouveaux Haemoproteidae parasites de chiroptères en relaxed clock. c. Topology acquired with the birth-death tree prior and a Thailande. Ann Parasitol Hum Comp 1984, 59:437-447. Polychromophilus murinus log-normal relaxed clock. The different haemosporidian clades are 12. Gardner RA, Molyneux DH: : a malarial parasite represented by the coloured triangles. The clade height represents the of bats: life-history and ultrastructural studies. Parasitology 1988, number of containing taxa. 96:591-605. 13. Dick CW, Patterson BD: Bat flies: obligate ectoparasites of bats. In Additional file 3: A topology rooted with Babesia provides little Micromammals and Macroparasites: From Evolutionary Ecology to information. The amino acid alignment provides too little contrast to Management. Edited by: Morand S, Krasnov BR, Poulin R. Tokyo: Springer- construct a tree with high support as most nodes are unsupported. A Verlag; 2006:179-194. very long branch separates the Babesia species from all Haemosporida. 14. Theodor O: An Illustrated Catalogue of the Rothschild Collection of Shown is the best tree of a ML analysis using a JTT + Γ + I substitution Nycteribiidae London: British Museum (Natural History); 1967. model and bootstrapping a 1,000 times. Closed dots: bootstrap value > 15. Corradetti A: Alcuni protozoi parassiti di Nycteribiidae del genere 90; Open dots: bootstrap values > 50. Listropoda. Ann Igiene 1936, 46:444-460. 16. Mer GG, Goldblum N: A haemosporidian of bats. Nature 1947, 159:444-444. 17. Petersen FT, Meier R, Kutty SN, Wiegmann BM: The phylogeny and evolution of host choice in the Hippoboscoidea (Diptera) as Acknowledgements reconstructed using four molecular markers. Mol Phylogenet Evol 2007, We would like to give a big thank you to Anne-Lyse Ducrest and Céline 45:111-122. Simon for advice on and assistance with the molecular work. We would also Witsenburg et al. Malaria Journal 2012, 11:53 Page 10 of 10 http://www.malariajournal.com/content/11/1/53

18. Megali A, Yannic G, Christe P: Disease in the dark: molecular Schmidt-Ott U, Kampmeier GE, Thompson FC, Grimaldi DA, Beckenbach AT, characterization of Polychromophilus murinus in temperate zone bats Courtney GW, Friedrich M, Meier R, Yeates DK: Episodic radiations in the revealed a worldwide distribution of this malaria-like disease. Mol Ecol fly tree of life. Proc Natl Acad Sci USA 2011, 108:5690-5695. 2011, 20:1039-1048. 19. Outlaw DC, Ricklefs RE: Rerooting the evolutionary tree of malaria doi:10.1186/1475-2875-11-53 parasites. Proc Natl Acad Sci USA 2011, 108:13183-13187. Cite this article as: Witsenburg et al.: The evolutionary host switches of Polychromophilus 20. Cummings MP, Meyer A: Magic bullets and golden rules: data sampling : a multi-gene phylogeny of the bat malaria genus in molecular phylogenetics. Zoology (Jena) 2005, 108:329-336. suggests a second invasion of mammals by a haemosporidian parasite. 21. Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S: MEGA5: Malaria Journal 2012 11:53. molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Mol Biol Evol 2011, 28:2731-2739. 22. Kück P, Meusemann K: In FASconCAT, version 1.0. Zool. Edited by: Forschungsmuseum A. Koenig, Germany; 2010:. 23. Guindon S, Gascuel O: A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol 2003, 52:696-704. 24. Stamatakis A: RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 2006, 22:2688-2690. 25. Nylander JAA: MrAIC.pl. Uppsala. Program Distributed by the Author Evolutionary Biology Centre, Uppsala University; 2004. 26. Ronquist F, Huelsenbeck JP: MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 2003, 19:1572-1574. 27. Drummond AJ, Ho SYW, Phillips MJ, Rambaut A: Relaxed phylogenetics and dating with confidence. PLoS Biol 2006, 4:699-710. 28. Drummond AJ, Nicholls GK, Rodrigo AG, Solomon W: Estimating mutation parameters, population history and genealogy simultaneously from temporally spaced sequence data. Genetics 2002, 161:1307-1320. 29. Drummond AJ, Rambaut A: BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol 2007, 7:214. 30. Kishino H, Hasegawa M: Evaluation of the maximum-likelihood estimate of the evolutionary tree topologies from DNA-sequence data, and the branching order in Hominoidea. J Mol Evol 1989, 29:170-179. 31. Jobb G: TREEFINDER Munich, Germany: Distributed by the author; 2008 [http://www.treefinder.de/], October 2008 edition. 32. Shimodaira H, Hasegawa M: Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol Biol Evol 1999, 16:1114-1116. 33. Yang ZH: PAML 4: phylogenetic analysis by maximum likelihood. Mol Biol Evol 2007, 24:1586-1591. 34. Christin P-A, Besnard G, Edwards EJ, Salamin N: Effect of genetic convergence on phylogenetic inference. Mol Phylogenet Evol . 35. Garnham PCC, Lainson R, Shaw JJ: A contribution to the study of Haematozoon parasites of bats - a new Haemoproteid, Polychromophilus deanei n. sp. Mem Inst Oswaldo Cruz 1971, 69:119-125. 36. Landau I, Chabaud AG, Miltgen F, Baccam D: Dionisia bunoi n. gen. n. sp. Haemoproteidae parasite du microchiroptère Hipposideros cyclops au Gabon. Ann Parasitol Hum Comp 1980, 55:271-280. 37. Perkins SL, Martinsen ES, Falk BG: Do molecules matter more than morphology? Promises and pitfalls in parasites. Parasitology 2011, 138:1664-1674. 38. Sanderson MJ, Shaffer HB: Troubleshooting molecular phylogenetic analyses. Annu Rev Ecol Syst 2002, 33:49-72. 39. Ane C, Larget B, Baum DA, Smith SD, Rokas A: Bayesian estimation of concordance among gene trees. Mol Biol Evol 2007, 24:412-426. 40. Heled J, Drummond AJ: Bayesian inference of species trees from multilocus data. Mol Biol Evol 2010, 27:570-580. 41. Duval L, Robert V, Csorba G, Hassanin A, Randrianarivelojosia M, Walston J, Nhim T, Goodman SM, Ariey F: Multiple host-switching of Haemosporidia parasites in bats. Malar J 2007, 6:157. Submit your next manuscript to BioMed Central 42. Morrison DA: Evolution of the Apicomplexa: where are we now? Trends and take full advantage of: Parasitol 2009, 25:375-382. 43. Valkiũnas G, Atkinson CT, Bensch S, Sehgal RNM, Ricklefs RE: Parasite misidentifications in GenBank: how to minimize their number? Trends • Convenient online submission Parasitol 2008, 24:247-248. • Thorough peer review 44. Ricklefs RE, Outlaw DC: A molecular clock for malaria parasites. Science • No space constraints or color figure charges 2010, 329:226-229. 45. Mattingly PF: The paleogeography of mosquito-borne disease. Biol J Linn • Immediate publication on acceptance Soc Lond 1983, 19:26. • Inclusion in PubMed, CAS, Scopus and Google Scholar 46. Wiegmann BM, Trautwein MD, Winkler IS, Barr NB, Kim JW, Lambkin C, • Research which is freely available for redistribution Bertone MA, Cassel BK, Bayless KM, Heimberg AM, Wheeler BM, Peterson KJ, Pape T, Sinclair BJ, Skevington JH, Blagoderov V, Caravas J, Kutty SN, Submit your manuscript at www.biomedcentral.com/submit